基于CO_2腐蚀形貌特征的腐蚀预测方法研究

采用灰度数据矩阵统计、小波变换及二值化等方法对不同油管钢经CO2腐蚀后的表面形貌图像进行特征提取。采用二值特征提取算法计算出以像素点个数表示的孔蚀面积,采用像素点集合求得蚀孔数目,并用能量灰度数据矩阵统计特征值,反映孔蚀表面腐蚀形貌凹凸起伏变化的复杂特征。结合多层前馈式反向传播BP神经网络,以腐蚀形貌图像的各向异性能量参数和小波变换后子图像的能量参数作为腐蚀类型判据,建立了基于BP神经网络的孔蚀速率诊断模型,诊断结果与实验结果基本吻合。

Study on Corrosion Prediction Method Based on CO2 Corrosive Morphology

The grey level data matrix statistics, wavelet transform and binarization were employed to extract surface morphology information from oil pipelines corroded by carbon dioxide. Using binary image extraction algorithm and pixel set, the pitting area expressed by numbers of pixels was obtained accurately, including the pitting number. The characteristic value of grey level data matrix statistics was taken to reflect the complexitiesof pitting surface. Having model was developed based of corrosion images and the with the testing results. multiplayer feed forward neural networks considered, a pitting velocity diagnosis on two corrosion type criterions, which including the anisotropic energy parameter image energy parameter after wavelet transform. The diagnosis results agree well